Context-Self Contrastive Pretraining for Crop Type Semantic Segmentation

Tarasiou, Michail; Güler, Rıza Alp; Zafeiriou, Stefanos

doi:10.1109/tgrs.2022.3198187

Cited by 10 publications

(5 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a similar spirit, [7] use a FPN [31] feature extractor, coupled with a CLSTM temporal model (FPN-CLSTM). The UNET3Df architecture [60] follows from UNET3D but uses a different decoder head more suited to contrastive learning. The U-TAE architecture [14] follows a different approach, in that it employs the encoder part of a UNET2D, applied on parallel on all images, and a subsequent temporal attention mechanism which collapses the temporal feature dimension.…”

Section: Related Work 21 Crop Type Recognitionmentioning

confidence: 99%

“…To accommodate a large set of experiments we only use fold 1 among the five folds provided in PASTIS. Finally, we use the T31TFM-1618 dataset [60] which covers a densely cultivated S2 tile in France for years 2016-18 and includes 20 distinct classes. In total, it includes 140k samples of size 48 × 48, each containing 14-33 acquisitions and 13 image bands.…”

Section: Training and Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

ViTs for SITS: Vision Transformers for Satellite Image Time Series

Tarasiou¹,

Chavez²,

Zafeiriou³

2023

Preprint

View full text Add to dashboard Cite

In this paper we introduce the Temporo-Spatial Vision Transformer (TSViT), a fully-attentional model for general Satellite Image Time Series (SITS) processing based on the Vision Transformer (ViT). TSViT splits a SITS record into non-overlapping patches in space and time which are tokenized and subsequently processed by a factorized temporo-spatial encoder. We argue, that in contrast to natural images, a temporal-then-spatial factorization is more intuitive for SITS processing and present experimental evidence for this claim. Additionally, we enhance the model's discriminative power by introducing two novel mechanisms for acquisition-time-specific temporal positional encodings and multiple learnable class tokens. The effect of all novel design choices is evaluated through an extensive ablation study. Our proposed architecture achieves state-of-the-art performance, surpassing previous approaches by a significant margin in three publicly available SITS semantic segmentation and classification datasets. All model, training and evaluation codes are made publicly available in https://github.com/michaeltrs/DeepSatModels.

show abstract

Section: Related Work 21 Crop Type Recognitionmentioning

confidence: 99%

Section: Training and Evaluationmentioning

confidence: 99%

ViTs for SITS: Vision Transformers for Satellite Image Time Series

Tarasiou¹,

Chavez²,

Zafeiriou³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…With the increasing prevalence of high-resolution remote sensing satellites in global Earth observation missions, high-resolution remote sensing data has become abundant and the primary source for Earth observation. Semantic segmentation of high-resolution remote sensing images [1][2][3][4][5] plays a vital role in understanding the distribution of ground object features, enabling refined urban management, environmental monitoring, natural resource assessment, crop analysis, precise surveying, and mapping. High-resolution remote sensing images possess distinct characteristics, including complex background information, dense targets, and rich ground object features.…”

Section: Introductionmentioning

confidence: 99%

SAINet: Spatially Adaptive Interaction Network for Semantic Segmentation of High-Resolution Remote Sensing Images

Song,

He,

Dai

et al. 2024

Preprint

View full text Add to dashboard Cite

Semantic segmentation of high-resolution remote sensing images provides detailed and precise feature information, playing a pivotal role in decision-making and analysis in sectors such as water management, agriculture, military, and environmental protection. However, most methods merely combine features from various branches directly, without a mechanism for spatial location feature screening, and treat all extracted features as equally important. To overcome these limitations, we introduce a novel spatially adaptive interaction network (SAINet) for dynamic interaction across different features in remote sensing semantic segmentation task. Specifically, we propose a spatial refined module that leverages local context information to filter spatial locations and extract salient regions. Following this, we introduce an innovative spatial interaction module that uses a spatial adaptive modulation mechanism to dynamically select and allocate spatial position weights. This facilitates interaction between local salient areas and global information, allowing the network to focus more on relevant regions. This flexible approach enables the network to capture more informative features, thereby enhancing segmentation accuracy. Experiments conducted on the DeepGlobe, Vaihingen, and Potsdam public datasets confirm the effectiveness and capability of SAINet. Our code and models will be publicly available

show abstract

“…Pixel-based classification tasks typically start by training a CNN classifier on small image patches, then using a sliding window method to predict the category of the central pixel [13][14][15], of which the drawback is that the trained network only predicts the central pixel of the input image, leading to low classification efficiency. Semantic segmentation, which aims to assign a specific class label to each pixel in an image with high process efficiency, is gradually gaining attention in the crop-mapping field [16]. For example, Zhang et al [17] combined a pyramid scene parsing network (PSPNet) [18] and GaoFen satellite images for cropland mapping.…”

Section: Introductionmentioning

confidence: 99%

Cross-Modal Segmentation Network for Winter Wheat Mapping in Complex Terrain Using Remote-Sensing Multi-Temporal Images and DEM Data

Wang,

Wu,

Gui

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

Winter wheat is a significant global food crop, and it is crucial to monitor its distribution for better agricultural management, land planning, and environmental sustainability. However, the distribution style of winter wheat planting fields is not consistent due to different terrain conditions. In mountainous areas, winter wheat planting units are smaller in size and fragmented in distribution compared to plain areas. Unfortunately, most crop-mapping research based on deep learning ignores the impact of topographic relief on crop distribution and struggles to handle hilly areas effectively. In this paper, we propose a cross-modal segmentation network for winter wheat mapping in complex terrain using remote-sensing multi-temporal images and DEM data. First, we propose a diverse receptive fusion (DRF) module, which applies a deformable receptive field to optical images during the feature fusion process, allowing it to match winter wheat plots of varying scales and a fixed receptive field to the DEM to extract evaluation features at a consistent scale. Second, we developed a distributed weight attention (DWA) module, which can enhance the feature intensity of winter wheat, thereby reducing the omission rate of planting areas, especially for the small-sized regions in hilly terrain. Furthermore, to demonstrate the performance of our model, we conducted extensive experiments and ablation studies on a large-scale dataset in Lanling county, Shandong province, China. Our results show that our proposed CM-Net is effective in mapping winter wheat in complex terrain.

show abstract

Context-Self Contrastive Pretraining for Crop Type Semantic Segmentation

Cited by 10 publications

References 42 publications

ViTs for SITS: Vision Transformers for Satellite Image Time Series

ViTs for SITS: Vision Transformers for Satellite Image Time Series

SAINet: Spatially Adaptive Interaction Network for Semantic Segmentation of High-Resolution Remote Sensing Images

Cross-Modal Segmentation Network for Winter Wheat Mapping in Complex Terrain Using Remote-Sensing Multi-Temporal Images and DEM Data

Contact Info

Product

Resources

About